Solar forecasting and variability analyses using sky camera cloud detection & motion vectors

Prediction of cloud location greatly increases the accuracy of solar generation forecasts when used in conjunction with regional meteorological data and historic data. Novel granular forecast techniques reduce intra-hour and minute-by-minute solar forecasting error by 50% compared to persistence models [3]. Our work creates cloud shadow maps on the ground to forecast power production ramps geospatially tagged and circuit topographically located to individual sites and aggregately to all sites on a distribution feeder. These forecasting data may be used to (1) reduce the detrimental impact on distribution systems due to voltage fluctuation caused by high PV penetration and (2) optimize utility operational and planning practices in the presence of high generation variability due to the presence of PV.